Human–GenAI Interaction in Higher Education: Trust as a Central Mechanism Shaping Learning Enhancement and Trade-offs

In the modern classroom, a new kind of tutor has arrived, one that never sleeps and can generate endless explanations, summaries, and ideas in seconds. This is generative artificial intelligence, a technology that has rapidly woven itself into the fabric of higher education. For students, it offers a way to tackle difficult concepts, draft essays, and organize thoughts with unprecedented speed. Yet, as these tools become ubiquitous, a quiet question has begun to trouble educators and researchers: does simply having access to this technology actually make students learn better? The prevailing assumption has often been that more use equals more learning, but the reality of how humans interact with machines is far more complex. The critical factor is not just the tool itself, but the relationship the student builds with it. This relationship hinges on trust—a psychological state where a learner believes the machine is reliable, truthful, and useful. Without trust, the tool remains a distant, suspicious object; with it, the tool becomes a partner in learning. However, this trust is a double-edged sword. It can unlock deeper engagement and confidence, but it can also lead to a dangerous overreliance where the student stops thinking for themselves. Understanding this delicate balance is essential for shaping how education evolves in the age of artificial intelligence.

A team of researchers at the SRM Institute of Science and Technology in India set out to map this uncharted territory. They wanted to move beyond simple questions about how often students use these tools and instead investigate the quality of the interaction and the role of trust in shaping learning outcomes. To do this, they gathered data from 237 higher education students, primarily undergraduates and postgraduates from commerce, science, and engineering fields in southern India. Through a detailed survey, they asked students to reflect on their experiences: how they interacted with generative AI, how much they trusted the information it provided, and how these factors influenced their academic performance and their own critical thinking skills. The researchers used a sophisticated statistical method to connect these dots, looking for patterns that revealed whether the interaction itself drove learning, or if something else was pulling the strings.

The study uncovered a clear and somewhat surprising truth: the mere act of using generative AI does not automatically lead to better learning. Instead, the key driver is the trust the student places in the system. The researchers found that when students interacted with the AI in a meaningful way, it fostered a sense of trust. This trust, in turn, was the primary engine that boosted learning outcomes. Students who trusted the AI reported feeling more motivated, more confident in their abilities, and more effective in their academic work. The interaction between the human and the machine mattered, but its power was almost entirely dependent on the student's belief in the machine's reliability. In fact, the direct link between simply using the tool and getting better grades was relatively weak. The real magic happened in the middle, where interaction built trust, and trust fueled learning.

However, the study also illuminated a significant trade-off that comes with this trust. The same mechanism that helped students learn more effectively also made them more likely to perceive risks and downsides. As trust in the AI grew, so did the students' awareness that they might be relying on it too much. The data showed a strong connection between high levels of trust and the feeling that they were sacrificing their own independent thinking. Students who trusted the AI deeply were more likely to report that they felt less inclined to solve problems on their own or engage in deep cognitive effort. This suggests that trust is not a purely positive force; it is a complex mechanism that simultaneously enhances learning and introduces the risk of overdependence. The students themselves voiced this tension in their written comments, noting that while the AI felt like a helpful tutor available at any time, it also had the potential to reduce their own initiative or take over their creativity.

The researchers also looked at whether these dynamics played out differently for different types of students. They found that the academic level and gender of the student mattered, but the frequency of AI use did not. Postgraduate students showed a stronger link between their interactions and the trust they developed, while undergraduate students seemed to benefit more from using the AI for specific, task-oriented jobs like writing or problem-solving. Interestingly, the amount of time a student spent using the AI had no significant impact on the structural relationships in the model. Whether a student used the tool a little or a lot, the quality of their interaction and the trust they built remained the deciding factors for their learning success. This finding challenges the common belief that simply increasing the volume of AI usage will improve education. Instead, it suggests that the depth of the relationship between the student and the technology is what truly counts.

Ultimately, this research paints a picture of generative AI in education not as a simple tool for efficiency, but as a partner in a complex psychological dance. The study confirms that the technology has the power to enhance learning, but only when it is approached with a calibrated sense of trust. If students trust the AI too little, they miss out on its benefits; if they trust it too much without critical oversight, they risk diminishing their own cognitive growth. The path forward for educators, therefore, is not to simply encourage more use, but to guide students in building a healthy, critical relationship with these systems. By focusing on the quality of interaction and fostering a trust that is grounded in understanding rather than blind reliance, higher education can harness the power of generative AI to create learning environments that are both effective and intellectually robust. The future of learning with AI depends less on the software itself and more on the human capacity to engage with it wisely.